sequential-read

What This Skill Does

The sequential-read skill transforms the way AI consumes long-form text by moving away from instant, monolithic summarization toward a genuine simulation of the reading experience. By breaking down documents—such as novels, academic papers, or dense technical reports—into semantic chunks, the agent builds a longitudinal record of evolving comprehension. It creates structured reflections at each step, capturing the reader's journey: tracking how opinions shift, verifying initial predictions, and logging answered questions. This results in a final synthesis that reflects a deep, grounded understanding of the source material as if it were processed over time.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/horace-claw/sequential-read Ensure that you have the necessary permissions for file-system access, as the skill utilizes local scripts to manage session state and metadata.

Use Cases

• Literary Analysis: Deeply engage with classic novels to track character development or thematic arcs with specific personas like 'literary critic'.
• Academic Research: Process long research papers to identify core arguments while noting how the author's methodology addresses the reader's initial skepticism.
• Technical Documentation: Read massive manuals or project specs, allowing the AI to 'learn' the system architecture as it parses through chunks, rather than just returning a surface-level summary.

Example Prompts

1. "/sequential-read ./books/the-great-gatsby.txt --lens literary-critic"
2. "/sequential-read ./papers/quantum-gravity-theory.pdf"
3. "/sequential-read show session_12345"

Tips & Limitations

• The Two-Phase Model: Do not be alarmed if the process appears to 'pause' or switch agents. The skill is designed to spawn a main reader for primary ingestion and a finisher agent for long documents. This ensures the model's context window remains optimized.
• Persistence: All sessions are saved locally via session_manager.py. You can list and revisit previous reading sessions at any time using the list and show commands.
• Limitations: The skill is intended for text-based ingestion. Large images or complex formatting within PDFs may require external OCR before reading. Ensure your file paths are absolute or relative to the OpenClaw execution directory to avoid access errors.